Linked optimal reactive contaminant source characterization in contaminated mine sites: case study

Application of a developed methodology for characterization of distributed contaminant sources in a contaminated mine site aquifer involving complex geochemical processes is presented. Linked simulation-optimization models are widely used as efficient tools for identifying unknown groundwater pollution sources. This is an essential first step in determining effective and reliable groundwater management and remediation strategies for a polluted aquifer. However, linking robust numerical models to simulate the transport processes of reactive chemical contaminant species in aquifers involving complex and highly nonlinear physical and geochemical process increases the computational burden extensively, and may affect the feasibility and efficiency of the methodology. To overcome this computational limitation and improve the computational feasibility by avoiding the necessity of repeated numerical simulations, genetic programming-based trained surrogate models are developed to approximately simulate such complex transport processes. Simulation model solution results are then approximated by training and testing genetic programming (GP)-based surrogate models. Performance evaluation of the GP models as surrogate models for the reactive species transport in groundwater demonstrates the feasibility of its use and the associated computational advantages. Some of the hydrogeologic parameter uncertainties are also incorporated in the surrogate models by using training samples obtained by random perturbation of the parameters. The linked simulation-optimization based methodology is evaluated for a contaminated former mining site in Queensland, Australia, as a limited case study